Hajime Kawahara

Thanks a lot!

Hi! Thanks. It makes things simple.

> We'll also merge https://github.com/radis/radis/pull/495 , which implements in Radis (and therefore in the API) the non-air broadening parameters Cool! Thanks a lot! @erwanp

OLA implemented in #283 and #284


Ah, I see. We can directly use `_calc_oa_lens()` to determine the optimal size of `block_size` because it does not need to be JITtable. I'll try to use it. Thanks!

You're right. L is much larger than I thought. Assuming the filter length of 50cm-1 @5000cm-1, the optimal block size is ~ 500cm-1. So, for instance, when we would compute...

 | scipy | Figure | [olaconv](https://github.com/HajimeKawahara/exojax/blob/presolar_initial/src/exojax/signal/ola.py) in ExoJAX | | ---- | ---- | ---- | | s1 | n L | input_length | | in1_step | L |...

Because I almost finished the implementation of PreMODIT, I restarted PreSOLAR algorithm, which aims to reduce the computational cost rather than device memory use. Also, it includes the cutoff of...

Breakdown of components tests - [x] olaform of lbd, nu_grid, shapefilter - [x] unbiased hat(lbd), hat(lbd) means the reshaped and padding array of lbd - [x] differentiable shape filter (i.e....